Impact of rapid on-site evaluation in endoscopic ultrasound-guided tissue acquisition for resectable pancreatic body and tail cancer

Abstract

BACKGROUND

Needle tract seeding is a potential complication of endoscopic ultrasound-guided tissue acquisition (EUS-TA), particularly in resectable pancreatic body and tail cancer. As the transgastric needle tract is generally not resected, reducing the number of punctures may be clinically important. We hypothesized that rapid on-site evaluation (ROSE) may help minimize the number of punctures while preserving diagnostic accuracy.

AIM

To investigate whether ROSE-guided EUS-TA is associated with a low number of punctures while preserving diagnostic accuracy in this setting.

METHODS

This multicenter retrospective study included 94 patients with resectable pancreatic body or tail cancer who underwent EUS-TA with ROSE at three tertiary centers from April 2020 to September 2024. ROSE assessed malignancy for each puncture in real time. The primary outcome was the number of punctures. Cytologic diagnostic performance, per-pass ROSE concordance, positive predictive value, and negative predictive value were evaluated. Factors associated with first-pass diagnostic discrepancy were analyzed using logistic regression.

RESULTS

A total of 94 patients were included. The mean number of punctures was 1.55 ± 0.80. Overall cytologic diagnostic accuracy was 96.8% on a per-patient basis. For per-pass specimens, concordance between ROSE and the corresponding final cytologic diagnosis was 86.0% (111/129); concordance was 97.8% (90/92) when ROSE indicated malignancy and 56.8% (21/37) when ROSE indicated no malignancy. Use of fine-needle biopsy needles was independently associated with a lower risk of diagnostic discrepancy.

CONCLUSION

ROSE-guided EUS-TA showed a low mean number of punctures and high diagnostic accuracy in resectable pancreatic body and tail cancer. Controlled studies are needed to determine whether ROSE reduces punctures.

Key Words: Endoscopic ultrasonography; Endoscopic ultrasound-guided fine-needle aspiration; Pancreatic cancer; Needle tract seeding; Rapid on-site evaluation; Puncture number

Core Tip: This multicenter retrospective study evaluated rapid on-site evaluation (ROSE) during endoscopic ultrasound-guided tissue acquisition for resectable pancreatic body and tail cancer, where needle tract seeding is a concern. The mean number of punctures was 1.55, and per-patient cytologic diagnostic accuracy remained high at 96.8%. A malignant ROSE result had a high positive predictive value, which may support termination of punctures when malignancy is identified on-site.

INTRODUCTION

Pathological confirmation of pancreatic cancer is routinely achieved using endoscopic ultrasound-guided tissue acquisition (EUS-TA), which has become the standard diagnostic approach for pancreatic tumors. However, recent studies have reported a non-negligible incidence of needle tract seeding (NTS) following EUS-TA in patients with resectable pancreatic body and tail cancer, with reported rates ranging from 0.857% to 3.4%[1,2]. This risk is of particular concern in patients with pancreatic body and tail cancer, because the needle tract created during transgastric EUS-TA is generally not removed at surgery[2,3].

From a theoretical standpoint, the risk of NTS is expected to increase proportionally with the number of needle passes. Therefore, minimizing the number of punctures during EUS-TA is especially important in patients with resectable pancreatic body and tail cancer. Despite this concern, the clinical significance of reducing the number of punctures during EUS-TA has not been sufficiently investigated[4-6].

Rapid on-site evaluation (ROSE) allows real-time assessment of malignancy during EUS-TA and may help avoid potentially unnecessary additional punctures[7-9]. However, previous studies of ROSE have mainly evaluated diagnostic accuracy or specimen adequacy on a per-patient basis in heterogeneous pancreatic masses. It remains unclear how ROSE performs as a real-time per-pass assessment of malignancy in resectable pancreatic body and tail cancer, where NTS is a particular concern. This multicenter retrospective study aimed to evaluate ROSE-guided EUS-TA in this specific setting, with a focus on the number of punctures and per-pass concordance between ROSE findings and the corresponding final cytologic diagnosis.

MATERIALS AND METHODS

Study design and patients

This was a multicenter, retrospective study conducted at three Japanese tertiary referral centers (Mie University Hospital, Ise Red Cross Hospital, and Shizuoka General Hospital). The study protocol was approved by the institutional review boards of all participating institutions. The approval number at Ise Red Cross Hospital was ER2024-123. The study was registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN000057987), and informed consent was obtained through an opt-out approach via institutional websites.

Patients were eligible for inclusion if they underwent EUS-TA with ROSE for resectable pancreatic body or tail cancer between April 2020 and September 2024, with curative-intent surgical resection planned at the time of EUS-TA. All included patients were ultimately pathologically diagnosed with invasive pancreatic ductal carcinoma, including adenocarcinoma, adenosquamous carcinoma, mucinous carcinoma, or anaplastic carcinoma.

Patients were excluded if detailed information regarding the EUS-TA procedure or ROSE was unavailable, or if consent for study participation was not obtained.

Definition of resectability and final diagnosis

Resectability was assessed using contrast-enhanced computed tomography before EUS-TA. Resectable pancreatic body and tail cancer was defined as the absence of distant metastasis and no tumor contact or invasion of major arteries or the portal venous system, in accordance with international consensus criteria[10]. Patients with borderline resectable pancreatic cancer were excluded.

The final diagnosis of invasive pancreatic ductal carcinoma was based on histopathological examination of surgical specimens whenever available. For patients who did not undergo surgical resection, clinical follow-up of more than 6 months was performed, and progressive disease consistent with pancreatic cancer on imaging and clinical course was required for the final diagnosis.

Tumor location was classified as pancreatic body or tail, with the left margin of the portal vein defined as the anatomical boundary. Operator experience was categorized as expert (≥ 150 EUS-TA procedures performed) or trainee (< 150 procedures). Procedure-related adverse events were assessed according to the American Society for Gastrointestinal Endoscopy lexicon[11].

EUS-TA procedures

All patients provided written informed consent for the EUS-TA procedure. EUS-TA was then performed using a linear-array echoendoscope (GF-UCT260; Olympus Medical Systems Corp., Tokyo, Japan, or EG-740UT/EG-580UT; FUJIFILM Medical Co., Ltd., Tokyo, Japan) under conscious sedation. Needle selection was left to the discretion of the endoscopist and included EZ Shot 3 Plus (Olympus Medical Systems Corp.), Acquire (Boston Scientific Corp., Marlborough, MA, United States), SonoTip TopGain (Medico’s Hirata Inc., Osaka, Japan), Trident (Century Medical, Inc., Tokyo, Japan), and SharkCore (Medtronic plc, Dublin, Ireland). All EUS-TA procedures in this study were performed using either 22-gauge or 25-gauge needles.

The basic procedural strategy was to terminate punctures once ROSE indicated malignancy, although the final number of punctures was determined by the endoscopist. Suction technique was selected at the discretion of the endoscopist and included no suction, standard syringe suction, or the slow-pull technique.

Specimen processing and ROSE

All obtained specimens were first expelled into a Petri dish and visually inspected. From each pass, the macroscopically most suitable portion of the obtained specimen, such as whitish or filamentous solid material, when present, was selected and compressed between two glass slides to prepare paired smears. One slide was stained with rapid May-Giemsa stain for ROSE, and the paired slide was stained with Papanicolaou stain; thus, both slides were prepared from the same specimen and contained identical cellular material.

ROSE was performed using rapid May-Giemsa-stained slides by experienced cytotechnologists. During ROSE, only the presence or absence of malignancy was assessed and communicated to the endoscopist. In some cases, comments regarding cellularity were additionally provided. In this study, “suspicious for malignancy” on ROSE was classified as malignant. Residual solid material not used for cytologic evaluation was fixed in formalin and processed into paraffin-embedded blocks for histological examination when available.

Final cytologic diagnoses were rendered independently for each pass specimen by experienced pathologists, primarily based on Papanicolaou-stained slides, with reference to May-Giemsa-stained slides when necessary. In addition to the presence or absence of malignancy, cytologic diagnoses included histologic subtyping. Histological diagnoses were made using hematoxylin and eosin staining. In this study, diagnoses of “suspicious for adenocarcinoma” were classified as malignant.

Outcome measurements

The primary outcome was the number of punctures required during EUS-TA. Diagnostic performance of cytology was evaluated on both a per-patient and per-specimen basis. Concordance between the preliminary diagnosis on ROSE and the final cytologic diagnosis of the corresponding specimen was assessed, and positive predictive value (PPV) and negative predictive value (NPV) of ROSE were calculated.

In addition, factors associated with discrepancy between ROSE and final cytologic diagnosis in the first-pass specimen were evaluated.

Statistical analysis

Continuous variables are presented as mean ± SD, and categorical variables are reported as n (%). Diagnostic accuracy of cytology was evaluated on a per-patient and per-specimen basis. Concordance between ROSE and final cytologic diagnosis was calculated, and PPV and NPV were determined. Logistic regression analysis was performed to identify factors associated with discrepancy between ROSE and final cytologic diagnosis in the first-pass specimen. Clinically relevant variables were included in the multivariate analysis using a stepwise selection method. A P value of < 0.05 was considered statistically significant. Statistical analyses were performed by the first author (Murabayashi T) using EZR (version 1.54; Saitama Medical Center, Jichi Medical University, Saitama, Japan).

RESULTS

Patient characteristics

Between April 2020 and September 2024, 94 patients who met the inclusion criteria underwent EUS-TA with ROSE for resectable pancreatic body or tail cancer at three institutions. All patients were considered surgical candidates at the time of EUS-TA and were finally diagnosed with invasive pancreatic ductal carcinoma. Baseline patient and tumor characteristics are summarized in Table 1. The mean age was 72.2 years, and 52 patients (55%) were male. Tumors were evenly distributed between the pancreatic body and tail (47 cases each). The mean lesion size, measured as the longest diameter on endoscopic ultrasound, was 19.7 ± 7.6 mm.

Table 1 Patient characteristics (n = 94), n (%).

Characteristic	Value
Age, mean ± SD (years)	72.2 ± 9.4
Sex, male/female	52/42
Tumor location
Body	47 (50)
Tail	47 (50)
Lesion size, mean ± SD (mm)	19.7 ± 7.6
Final diagnosis
Adenocarcinoma	92 (98)
Adenosquamous carcinoma	1 (1.1)
Anaplastic carcinoma	1 (1.1)
Surgical resection	84 (89)

Procedural characteristics and diagnostic performance

Procedural details and outcomes of EUS-TA are shown in Table 2. EUS-TA was performed by trainees in 55 cases (59%). Fine-needle biopsy (FNB) needles were used in 63 patients, whereas fine-needle aspiration (FNA) needles were used in 31 patients; 22-gauge and 25-gauge needles were used in 54 and 40 patients, respectively. The mean number of punctures was 1.55 ± 0.80. A single puncture was performed in 56 patients (60%), whereas two, three, and four punctures were performed in 28 (30%), 6 (6.4%), and 4 patients (4.3%), respectively.

Table 2 Characteristics and outcomes of endoscopic ultrasound-guided tissue acquisition (n = 94), n (%).

Variable	Value
Endoscopist performing EUS-TA
Trainee	55 (59)
Expert	39 (41)
Needle used
Franseen, 22-gauge	18 (19)
Franseen, 25-gauge	13 (14)
Menghini, 22-gauge	20 (21)
Menghini, 25-gauge	11 (12)
Trident, 22-gauge	11 (12)
Trident, 25-gauge	13 (14)
Fork-tip, 22-gauge	5 (5.3)
Fork-tip, 25-gauge	3 (3.2)
Suction technique
Suction with syringe	65 (71)
Slow-pull	18 (20)
No suction	9 (9.6)
Unknown	2 (2.1)
Number of punctures
1	56 (60)
2	28 (30)
3	6 (6.4)
4	4 (4.3)
Diagnostic accuracy of cytology (per patient)	91/94 (96.8)
Diagnostic accuracy of histology (per patient)	62/69 (89.9)
Adverse events	0

EUS-TA: Endoscopic ultrasound-guided tissue acquisition.

The overall diagnostic accuracy of cytology based on EUS-TA was 96.8% (91/94) on a per-patient basis. When analyzed per specimen, the diagnostic accuracy of cytology across all punctures was 81.8% (108/132). Histological evaluation was additionally performed in 69 patients, yielding a diagnostic accuracy of 89.9% (62/69); however, histology did not provide an incremental diagnostic benefit over cytology alone. No procedure-related adverse events were observed.

Ultimately, 84 patients (89%) underwent surgical resection, with histopathology confirming pancreatic cancer in all cases. The remaining 10 patients showed clinical progression consistent with pancreatic cancer.

Diagnostic yield according to ROSE and number of punctures

Figure 1 illustrates the diagnostic flow based on ROSE findings across sequential punctures. Malignancy was identified on ROSE during the first puncture in 69 patients (73%). Among these patients, 13 underwent additional punctures at the discretion of the endoscopist, primarily to obtain a larger tissue volume despite an initial malignant diagnosis on ROSE. In 25 patients (27%), malignancy was not identified on ROSE during the first puncture, prompting additional punctures.

Figure 1 Flow diagram of diagnostic outcomes according to rapid on-site evaluation during endoscopic ultrasound-guided tissue acquisition. ¹Final cytologic diagnosis of the corresponding pass. ²Patients who underwent a second pass despite malignancy being identified on rapid on-site evaluation during the first pass. ROSE: Rapid on-site evaluation; EUS-TA: Endoscopic ultrasound-guided tissue acquisition.

Malignancy was subsequently identified on ROSE during the second, third, or fourth puncture in most of these patients. ROSE was performed in a total of 129 specimens. Of these, ROSE indicated malignancy in 92 specimens, and malignancy was confirmed on final cytologic diagnosis in 90 of those specimens. ROSE indicated non-malignancy in 37 specimens; among these, 21 specimens were also non-malignant on final cytologic diagnosis, whereas 16 were malignant. Overall, the concordance rate between the preliminary diagnosis on ROSE and the final cytologic diagnosis of the corresponding specimen was 86.0% (111/129). The PPV of a malignant diagnosis on ROSE for malignancy on final cytologic diagnosis was 97.8% (90/92), whereas the NPV of a non-malignant diagnosis on ROSE for non-malignancy on final cytologic diagnosis was 56.8% (21/37).

Factors associated with discrepancy between ROSE and final cytologic diagnosis

Thirteen patients showed discrepant results between the preliminary diagnosis on ROSE and the final cytologic diagnosis in the first-pass specimen (Figure 1). Univariate and multivariate logistic regression analyses were performed to identify factors associated with this discrepancy. On univariate analysis, only the use of a FNA needle was significantly associated with diagnostic discrepancy, whereas tumor location in the pancreatic body reached significance only on multivariate analysis (Table 3).

Table 3 Factors associated with discrepancy between rapid on-site evaluation and final cytologic diagnosis.

	Univariate analysis			Multivariate analysis
	OR	95%CI	P value	OR	95%CI	P value
Tumor location (body)	2.55	0.725-8.94	0.145	5.30	1.19-23.7	0.0291
Lesion size < 20 mm	1.46	0.450-4.72	0.529
Trainee operator	0.802	0.247-2.60	0.713
FNA needle	6.03	1.69-21.6	0.0058	7.85	1.94-31.8	0.0129
25-gauge needle	1.18	0.365-3.84	0.777
Suction with syringe	5.30	0.649-43.2	0.120

Multivariate analysis was performed using a stepwise selection method. OR: Odds ratio; CI: Confidence interval; FNA: Fine-needle aspiration.

DISCUSSION

In this multicenter retrospective study, we evaluated the impact of ROSE on the number of punctures during EUS-TA in patients with resectable pancreatic body and tail cancer. The key finding was that ROSE-guided EUS-TA was performed with a low mean number of punctures while maintaining high diagnostic accuracy. Specifically, the preliminary ROSE diagnosis and final cytologic diagnosis in the first-pass specimen were concordant in 86% of patients. This high first-pass concordance suggests that ROSE has the potential to help optimize the number of punctures by supporting real-time decisions regarding additional sampling.

In the present cohort, a mean of 1.55 punctures per patient resulted in an overall cytologic diagnostic accuracy of 96.8%. In hypothetical fixed-pass strategies applied to this cohort, the diagnostic accuracy would be 83.0% (78/94) with one puncture, 93.6% (88/94) with two punctures, and 95.7% (90/94) with three punctures. Thus, this exploratory comparison suggests that ROSE has the potential to reduce unnecessary additional punctures while maintaining diagnostic accuracy. However, this finding should be interpreted cautiously because this study did not include a non-ROSE control group.

The high PPV of ROSE observed in this study (97.8%) supports its role as a reliable indicator for terminating tissue acquisition when malignancy is identified on-site. In contrast, the NPV was relatively low (56.8%), indicating that a non-malignant preliminary diagnosis on ROSE does not reliably exclude malignancy on final cytologic evaluation. This discrepancy may reflect a cautious diagnostic approach by cytotechnologists, who may avoid overcalling malignancy in equivocal specimens during ROSE. From a clinical perspective, these results suggest that while a malignant ROSE diagnosis can justify termination of punctures, repeated punctures until a malignant ROSE diagnosis is obtained may not always be necessary, particularly when the number of punctures is increasing.

Numerous previous studies[7,8,12-15] have evaluated the role of ROSE during EUS-TA for pancreatic masses; however, most of these investigations focused on per-patient diagnostic accuracy or sensitivity for malignancy, rather than on pass-by-pass diagnostic performance. In many prior studies, ROSE was evaluated as a per-patient diagnostic tool, rather than as a real-time assessment of malignancy for each individual puncture. Therefore, the clinical value of ROSE as a real-time, per-pass tool for deciding whether additional punctures are necessary remains unclear. To our knowledge, this is the first study to systematically evaluate concordance, PPV, and NPV between the ROSE diagnosis and the final cytologic diagnosis on a per-pass basis, with a specific focus on malignancy. By analyzing ROSE findings for each puncture in relation to the corresponding final cytologic diagnosis, this study provides new information on how ROSE may support real-time decisions regarding whether to continue or terminate punctures during EUS-TA.

Eloubeidi et al[16] reported a concordance rate of 98% between ROSE and the final cytologic diagnosis in a large series of EUS-TA procedures, including 656 lesions with 229 pancreatic masses, using a cumulative, per-patient assessment based on multiple puncture specimens. However, their analysis did not evaluate ROSE performance on a pass-by-pass basis, and the reported concordance reflects the final integrated result of repeated sampling within each case. Consequently, their methodology differs fundamentally from that of the present study and does not address the role of ROSE in guiding real-time decisions to terminate or continue punctures.

Previous studies and meta-analyses conducted more than a decade ago reported that ROSE improved specimen adequacy and diagnostic accuracy in EUS-TA for pancreatic masses[7,8]. In contrast, more recent randomized controlled trials have shown limited additional benefit of ROSE in the era of FNB needles, particularly with respect to diagnostic accuracy[13-15]. Notably, these trials included heterogeneous populations encompassing both pancreatic head and body/tail tumors, as well as resectable and unresectable disease. The novelty of the present study is its focus on resectable pancreatic body and tail cancer, where NTS is a particular concern, and its per-pass evaluation of ROSE findings. In this specific setting, our findings suggest that ROSE may remain clinically useful as a real-time decision-making tool during EUS-TA, rather than simply as a method for improving overall diagnostic accuracy.

Regarding needle type, the use of FNB needles was independently associated with a lower risk of discrepancy between ROSE and the final cytologic diagnosis. This finding likely reflects improved specimen quality obtained with FNB needles. In contrast, needle gauge (25 G vs 22 G) was not associated with diagnostic performance or discrepancy between ROSE and the final cytologic diagnosis. Given the comparable diagnostic results between 25 G and 22 G needles, the use of thinner needles may be a reasonable option, as they could theoretically reduce the risk of NTS.

Several limitations of this study should be acknowledged. First, this was a retrospective, single-arm study, and therefore a direct comparison between ROSE-guided and non-ROSE-guided strategies was not possible. Thus, it cannot be concluded from this study that ROSE reduced the number of punctures. The observed low mean number of punctures suggests that ROSE may help avoid unnecessary additional punctures, but this possibility should be confirmed in further controlled studies. Second, the number of punctures and decisions regarding additional sampling were determined at the discretion of the endoscopist and were not standardized. Third, the interpretation of ROSE depends on the expertise of the cytotechnologist, which may limit the generalizability of our findings.

CONCLUSION

In conclusion, ROSE-guided EUS-TA was performed with a low mean number of punctures and high diagnostic accuracy in patients with resectable pancreatic body and tail cancer. Further controlled studies are needed to determine whether ROSE truly reduces the number of punctures compared with non-ROSE strategies.